AI-optimized skill for building agents with LangGraph (>=1.0).
Structured reference covering the full LangGraph surface — from first install to production deployment — organized for AI coding assistants (Claude Code, Cursor, GitHub Copilot) and human learners alike.
Companion project: deepagents-skill — multi-agent systems with Deep Agents + LangGraph hybrid architectures.
langgraph-skill/
├── 00_Get_started/ # Overview, install, quickstart, mental model
├── 01_Capabilities/ # Persistence, streaming, interrupts, memory, subgraphs
├── 03_Production/ # App structure, testing, LangSmith, deployment
├── 04_LangGraph_APIs/ # Graph API vs Functional API — choosing & using
├── ROUTER.md # Decision tree — start here
└── LICENSE
For AI assistants: point the assistant at ROUTER.md — it contains a decision tree that routes to the exact file needed for a given task.
For humans: browse 00_Get_started/ to get started, or read ROUTER.md for the full map.
This is a v0 release — structured reference documentation, not parameterized executable skills. The agent reads, understands, and decides how to act. This works well with strong models (Opus, Sonnet) but requires good instruction-following capability.
- Add parameterized skill entry points (accept intent, output code)
- Keep content in sync with LangGraph releases
- Add runnable code templates (like deepagents-skill's
templates/) - Community-contributed patterns and examples
Contributions welcome — see Contributing.
- Fork the repo
- Create a branch (
git checkout -b feature/my-skill) - Follow the existing file naming and structure conventions
- Submit a PR with a clear description of what you added or changed